Impact of macro-structural reforms on the productivity...
Transcript of Impact of macro-structural reforms on the productivity...
1
Centre for Globalization Research School of Business and Management
Impact of macro-structural reforms on the productivity growth of regions: distance to the
frontier matters
CGR Working Paper 86
Sabine D’Costa Enrique Garcilazo
Joaquim Oliveira Martins
Abstract Using a panel of 265 regions from 24 OECD countries from 1997 to 2007, we explore the impact of nation-wide macroeconomic and structural policies on the productivity growth of subnational regions. We find that average relationships between nation-wide policies and the growth of regions can hide strong differentiated effects according to the distance to the frontier: relaxing employment protection legislation on temporary contracts, lowering barriers to trade and investment as well as increasing trade openness enhances productivity growth in lagging regions, whereas reducing barriers to entrepreneurship or higher levels of government debt has a positive effect on regions that are closer to the productivity frontier. Keywords: structural reforms; regional growth; lagging regions. JEL codes: R11, R58, O18
http://www.busman.qmul.ac.uk/cgr
CG
R W
orki
ng P
aper
Ser
ies
1
Impact of macro-structural reforms on the productivity growth of regions: distance to the frontier matters
Sabine D’Costa a Enrique Garcilazo b
Joaquim Oliveira Martins b
a University of Westminster, London, UK and SERC b OECD1, 2 rue André Pascal, 75116 Paris, France
Abstract:
Using a panel of 265 regions from 24 OECD countries from 1997 to 2007, we explore the impact of nation-wide macroeconomic and structural policies on the productivity growth of subnational regions. We find that average relationships between nation-wide policies and the growth of regions can hide strong differentiated effects according to the distance to the frontier: relaxing employment protection legislation on temporary contracts, lowering barriers to trade and investment as well as increasing trade openness enhances productivity growth in lagging regions, whereas reducing barriers to entrepreneurship or higher levels of government debt has a positive effect on regions that are closer to the productivity frontier.
Keywords: structural reforms; regional growth; lagging regions.
JEL codes: R11, R58, O18.
1 The views expressed are those of the authors and do not reflect those of the OECD or its Member countries.
2
1. Introduction
This paper aims at analysing the impact of macro-structural factors on the productivity
growth of regions. These economy-wide factors may be defined in a uniform way for all
regions in a given country, but their impact region by region can be very asymmetric. In this
way, we aim at bridging the gap between the national and the regional dimensions in the
study of economic growth, as well as in economic policy. On the one hand, at the country
level, the neoclassical literature investigating economic growth based on Cobb-Douglas
production functions (Solow, 1956 and Swan, 1956) which evolved towards the endogenous
growth framework (Lucas, 1988 and Romer, 1986) focused on understanding the country-
level drivers of national growth, including country-level policies. On the other hand, the
literature investigating economic growth dynamics at the regional level explains regional
growth based on regional factors. This includes studies of regional convergence (Sala-i-
Martin, 1996, Barro and Sala-i-Martin, 2004) and more recently models of regional growth
using frameworks from the New Economic Geography (Minerva and Ottaviano, 2009,
Desmet and Rossi-Hansberg, 2010)2 . There is still a gap between these two bodies of
literature and as a result the link between country-level factors and regional economic growth
needs to be better understood. Recent work has advanced our understanding of how the
regional dimension maps into and contributes to aggregate growth (see OECD, 2011). Che
and Spilimbergo (2012) using a limited set of structural reform indicators analysed how they
impact regional convergence; however the study of how country-wide factors influence
performance at the regional level is still nascent.
2 See Breinlich et al. (2013) and OECD (2009) for a review.
3
The country-level factors we examine comprise a broad package of structural policies,
including product market regulation and labour market legislation, as well as macroeconomic
factors such as trade exposure, the level of inflation and government debt. We consider their
effect on the productivity growth of OECD regions, measured as growth in GDP per worker.
In particular we seek to explore how this impact might vary across regions depending on their
productivity gap with the most productive region in their country (the frontier region). We
also examine the pass-through effect by estimating how productivity growth in the frontier
region affects productivity growth in other regions within a country and the catching-up effect
by estimating whether regions that are farther from the frontier (lagging regions) grow faster.
The fact that we are using regional data enables to control for fixed-effects without creating
the usual collinearity problems with country-wide policy variables, which also may also
enable to estimate better the effects of these policies.
We believe our work carries important conclusions for both macro-structural and
regional policies. Regional policy has evolved over the past decades from a paradigm
dominated by temporary subsidies and short-term corrections in regional imbalances to a
modern approach focusing on competitiveness and growth with an aim to boost the overall
performance of countries.3 A criticism of EU regional transfers has been made for example in
Boldrin and Canova (2001) who find that productivity in poorer regions was unaffected by the
amount of transfers received. The authors conclude that these policies simply have a
redistributive role without enhancing either aggregate growth or that of lagging regions in the
EU. More recently, Breidenbach et al. (2016) found that the disbursement of EU structural
funds is negatively correlated with regional growth and does not seem to contribute
3 See McCann and Ortega-Argiles (2013) for a detailed description of the changes to EU cohesion policy for example.
4
effectively to foster income convergence across regions. Other work on EU regional transfers
has highlighted the fact that their overall efficiency could be improved by reallocating these
transfers towards a select group of target regions (Becker et al., 2012). Our paper aims to
contribute to the current debate over the conditionality of regional transfers whereby
unconditional transfers have been deemed insufficient and transfers are now increasingly tied
to specific structural improvements.4 Our results indicate that lagging regions differ from
leading regions in their response to structural reforms and therefore that regionally
differentiated policies could have a significant role to play in national growth policy. We find
that average relationships between nation-wide policies and the growth of regions can hide
strong differentiated effects according to the distance to the frontier: relaxing employment
protection legislation on temporary contracts, lowering barriers to trade and investment as
well as increasing trade openness enhances productivity growth in lagging regions, whereas
reducing barriers to entrepreneurship or higher levels of government debt has a positive effect
on regions that are closer to the productivity frontier.
The paper is structured around six sections. In the next section we provide an
overview of the literature and of our conceptual framework. The third section describes the
model and Section 4 is dedicated to the data. Section 5 presents the results of our estimates as
well as robustness checks. The final section presents our conclusions.
4 The heterogeneity of the effect of transfers on regional growth is analysed in Becker et al. (2013), who find that regions with better governance and human capital enjoy greater growth in response to EU transfers.
5
2. Conceptual framework and review of the literature
Our empirical model is inspired by the model in Bourlès et al. (2013). This is a
version of the neo-Schumpeterian endogenous growth model by Aghion et al. (1997), which
highlights the costs of market restrictions (e.g. regulatory barriers) in upstream sectors have
an impact on the productivity growth of downstream sectors. The conclusion of their model is
that the (induced) lack of competition in upstream sectors leads to lower productivity growth
in downstream sectors.
In addition, Bourlès et al. (2013) shows the relevance of two factors that have been
identified in the endogenous growth literature as influencing positively sector productivity
growth. First, growth at the international technological frontier for a given sector has a
positive effect on growth in lagging country-sectors: this is called technological pass-through.
Second, by a catching-up effect, the efficiency gap between this frontier sector and the
follower sectors also enhances growth in the follower sectors. As highlighted in Acemoglu et
al. (2006) and Aghion and Howitt (2006), productivity growth in follower countries increases
in the productivity growth at the world technological frontier (pass-through effect) owing to
technological spillovers from the frontier and also increases in the productivity gap between
the follower country and the technological frontier owing to technological adoption (catching-
up effect). A result of this is that regulations that curb competition between firms will reduce
the incentive for firms to adopt the technology available at the frontier. This will slow down
the catching-up of follower countries that are far from the technological frontier. Therefore
the speed of catching-up depends not only on the distance to the frontier, but also indirectly
on regulations or policies that might affect the distance to the frontier (Bourlès et al., 2013).
Another consequence is that, as shown in Griffith et al. (2004), countries that invest in R&D
6
benefit from a double productivity boost, through greater catching-up possibilities owing to
innovation and also through easier technological adoption from the frontier.
The structural policies that our and previous papers consider can affect productivity
through two separate and opposing effects, the escape entry effect and the discouragement
effect.5 In the context of deregulation leading to heightened competition between firms, the
escape entry effect would prevail in countries with a significant proportion of firms with
productivity levels near the productivity frontier, and the discouragement effect would prevail
in countries where most firms are far away from the frontier. In that case, in follower
countries far from the frontier, the discouragement effect would prevent innovation taking
place and reduce catching-up. Conversely, where the escape entry effect prevails, competition
will increase average productivity. This effect may in fact prevail too in lagging countries or
regions far from the frontier, where higher domestic or foreign competition may increase
firms’ incentives to adopt better technologies.
Recent work on the causes of the generalised productivity slowdown in OECD
countries has also highlighted that there is an apparent average decline of productivity
catching-up at the firm and sectoral level (OECD, 2015). This stylised fact is also confirmed
when observing productivity trends in frontier and lagging regions (OECD, 2016). Among the
drivers that could promote or hinder catching-up, OECD (2015) stresses the relative openness
of economies to trade and investment, with an essential component associated with the
participation in global value-added chains (GVCs); the existence of barriers that generate
disincentives for up-scaling of firms, which in turn prevent firms from reaching the size
required to face the fixed costs to adopt innovation or enter new markets; and finally, the
5 See Aghion et al. (2004, 2006).
7
existence of market restrictions that may also prevent the necessary firm selection and
reallocation mechanisms from fully operating.
We transpose the country-sector framework of Bourlès et al. (2013) at the regional
level, examining the impact of nation-wide structural policies and macroeconomic factors on
regional productivity growth, while simultaneously examining the pass-through and the
catching-up effects. For the former effect we determine whether regional productivity growth
increases with the growth of the country’s frontier region (i.e. the region with the highest
productivity level in its country in a given year) and for the latter we determine whether
regional growth increases in distance to the frontier region.
Shifting the analysis from the sectoral level to the region level of course means that
the relevant explanatory variables present in the growth regressions that we undertake will be
different. In the rest of this section, we explain our choice of explanatory variables.
Our paper is concerned with the relationships between economic growth and levels of
product market regulation, which have been studied before at the country level by Nicoletti
and Scarpetta (2003) for OECD countries. The results in Nicoletti and Scarpetta (2003)
indicate that differences in regulatory reform explain part of the cross-country growth
disparities. They find that lower levels of entry barriers and state control enhance productivity
growth, particularly in countries that lag behind in technology adoption. They also find
productivity gains from privatization. This leads us to ask whether such reforms might explain
differences in the growth of subnational regions. In addition, we also consider the impact of
employment protection legislation (EPL) on regional growth.
Our approach is also related to that of Bassanini and Scarpetta (2001), who build a
country-level “policy-augmented” growth model analysing the effects of macroeconomic
8
policies such as inflation targeting, fiscal policy or international trade on national economic
growth in OECD countries. Their findings suggest that high inflation hinders output growth,
possibly due to its negative effect on investment and capital accumulation. They also find a
negative effect of the size of government on growth, whereas trade exposure is found to be
positively associated with output growth.
As highlighted in Bassanini and Scarpetta (2001), government deficit can affect
country-level growth by reducing private sector investment, and by implying a level of
taxation that changes the efficient allocation of resources in the economy. In spite of the
positive effects of public spending, medium to high levels of deficit tend to curb economic
growth. The magnitude of the effect depends on the type of financing of the deficit (i.e. how
distortionary the taxes are) and the type of public investment undertaken (i.e. how productive
it is). At the regional level, in contrast to Bassanini and Scarpetta (2001), by estimating the
effect of local taxes and public expenditures on regional economic growth in Korea, Kim
(1997) finds that overall the positive effect of local government investment on regional
growth outweighs the negative effect of local taxes. Rodriguez-Pose and Fratesi (2003) also
find a small positive impact of European structural funds on regional growth in the EU. As
regional policy funds are typically targeted on gaps of regional GDP per capita, depending on
the distance to the frontier, the balance between positive and negative effects may change.
International trade can also enhance economic growth, by reinforcing the efficient
allocation of resources according to patterns of comparative advantage, by increasing the
scale of production, facilitating the flow of technologies and knowledge, and increasing levels
of competition. The New Economic Geography and growth literature, in particular Martin and
Ottaviano (1999) suggest there is a permanent effect of trade integration on economic growth.
9
In contrast, Minniti and Parello (2011), using a spatial model of endogenous growth, predict
that trade integration has only a short term impact on growth, which is positive when there are
positive R&D spillovers. In terms of empirical evidence, Sachs et al. (2002), aiming to
explain the differences in economic performance across Indian states, find that after the
reforms of 1991 the surge in international trade has been a positive factor of growth.
Concerning inflation, despite evidence of its growth effects at the country level
(Bassanini and Scarpetta, 2001), no clear predictions can be made on its effects on regional
productivity. We include it more as control variable given its importance as one of the
macroeconomic stabilisation indicators.
Among the structural policies we include in our analysis are reforms aimed at
lowering employment protection legislation, state control, barriers to entrepreneurship and
barriers to trade and investment. All these structural variables may affect regional
productivity in a differentiated way depending on the distance to the frontier.
From the outset, it could be noted that the majority of the frontier regions in the
OECD contain large cities, whereas most lagging regions are located in intermediate and
predominantly rural areas (OECD, 2016). In this context, lagging regions may be particularly
sensitive to barriers to entrepreneurship precisely because they do not fully benefit from
density and agglomeration effects associated with the presence of other firms allowing for
specialisation or knowledge spillovers. Indeed, Stephens et al. (2013), focusing on lagging
rural regions of the Appalachians in the USA, find that, in such regions, employment growth
is better enhanced by the presence of creative workers or a tradition of self-employment and
entrepreneurship in the area than by knowledge-based factors such as the presence of
universities, a greater proportion of patents or higher high-technology employment shares.
10
The openness of international trade and investment may also affect regions differently.
OECD (2016) finds that exposure to tradable sectors appears to be one of the main drivers of
the regional productivity catching-up in lagging regions. A possible explanation is that in
tradable sectors (mainly industry), the international competitive pressures generate a sort of
unconditional convergence (Rodrik, 2013), which is less dependent on specific country or
regional conditions. The barriers to trade and investment could then be particularly
detrimental for the productivity performance of the lagging regions.
State control can have different effects on regional growth depending on the form it
takes and on the distance to the frontier. The extent of state ownership with direct control on
public employment and investment can be a strong factor of growth in regions farther away
from the frontier. On the other hand, state involvement in business operations, such as price
controls and regulation can also reduce the efficient allocation of resources and affect
incentives to innovate, with a detrimental impact on growth in frontier regions.
High levels of employment protection legislation (EPL) can have a limiting effect on
competition and on creative destruction by reducing the flexibility of firm size: this can
restrict up-scaling of firms but also limit the exit or downsizing of inefficient firms. By
lowering the quality of job-market matching, it can also prevent effective resource
reallocation with a detrimental effect on aggregate productivity growth. Lowering restrictions
on temporary contracts can have a stronger effect on growth for regions that are farther away
from the frontier, as these generally have thinner labour markets and are therefore more
vulnerable to labour market rigidities.
We limit our analysis to the main structural reform and macroeconomic indicators.
This is both motivated by the existence of institutional data and composite indicators that
11
allow for a comparative analysis across countries and, also, by the fact that these macro-
structural variables have been used to implement ex-ante conditionalities associated with EU
regional policy funds.
Our empirical framework focuses on pass-through, catching-up and structural
framework conditions. It also accounts for time-invariant regional drivers of growth. It does
not explicitly control for other time-varying region-level drivers of regional growth such as
physical and human capital and innovation. Although at the country level both the
neoclassical theory of growth (starting with Solow, 1956 and Swan, 1956) and endogenous
growth models (Romer, 1986 and 1990, Lucas, 1988 and Aghion and Howitt, 1998)
emphasise the role of physical and human capital accumulation on economic growth, the
evidence at the regional level is mixed. Chandra and Thompson (2000) and Michaels (2008)
find that improved access to interstate highways in rural US counties increased firm earnings
and Duranton and Turner (2011) find that population growth in US Metropolitan Statistical
Areas responds positively to increases in the road network. However Crescenzi and
Rodriguez-Pose (2012) fail to find a role for transport infrastructure in regional economic
growth among EU regions. Turning to human capital, although empirical evidence on the
importance of human capital for regional growth can be found in Glaeser et al. (1995),
Henderson et al. (1995) and Rauch (1993), we do not find such evidence in our dataset. This
may be due to the fact that investment in education has been relatively uniform across
countries, at least at the TL2 level, thus it cannot really discriminate between the fast and slow
growing regions within countries. Similarly, we do not find in our data a relationship between
innovative activity and regional productivity growth.6
6 For evidence on the convergence of R&D policies across regions see OECD (2016).
12
Our paper is part of a limited set of studies that have combined the notions of
macroeconomic policies, regional economic growth and convergence. Studying the case of
Indian states, Ahluwalia (2000) discusses the reasons for inter-state differences in economic
growth in India in the 1990s. He finds that variations in growth are best explained by
variations in certain state-specific characteristics. However, he does not provide empirical
evidence on how nation-wide structural reforms can have different impacts on states as these
differ in their characteristics. More recently, related work by Che and Spilimbergo (2012) has
estimated the effect of structural reforms on the speed of convergence between regions in
developed and developing countries, concluding that financial development, trade openness,
sound institutions and some labour market reforms favour regional convergence.
3. Description of the model
As previously explained, our model is based on a modified version of Bourlès et al.
(2013) adapted to the regional context. Our policy-augmented growth model is based on a
regional production function rather than a national one, as is often done in the regional growth
literature. It allows us to estimate the effects of macroeconomic and structural policies on
regional productivity growth and simultaneously measure how this effect varies with respect
to a region’s distance to the “frontier” or most productive region in the country (the catching-
up effect) and the direct impact of the frontier region on productivity growth (the pass-
through effect). Our hypothesis is that regional productivity growth is positively related to the
productivity growth of the frontier region within the country and positively related to the
productivity gap with the frontier region (in other words productivity growth increases with
distance to the productivity frontier as lagging regions catch up).
13
In terms of structural policies, we consider labour market legislation, the level of state
control, barriers to entrepreneurship and barriers to trade and investment, which are described
in the next section. Our macroeconomic variables are trade exposure, government debt and
inflation.
We estimate the following reduced form equation:
∆ ln𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡 = 𝛽𝛽1∆ ln𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹𝐶𝐶 ,𝑡𝑡 + 𝛽𝛽2 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1 + 𝛽𝛽3𝑃𝑃𝑃𝑃𝑃𝑃𝐶𝐶,𝑡𝑡−1 +
𝛽𝛽4𝑃𝑃𝑃𝑃𝑃𝑃𝐶𝐶,𝑡𝑡−1 .𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1 + 𝛾𝛾𝑟𝑟 + 𝜁𝜁𝑡𝑡 + 𝜀𝜀𝑟𝑟𝑡𝑡 (1)
∆ ln𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡 is the percentage growth in region r’s productivity between t-1 and t
(productivity measured as GDP per worker), ∆ ln𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹𝐶𝐶,𝑡𝑡 is the percentage growth of the
country frontier region’s productivity in year t and 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1 is the lagged productivity
gap with the country’s frontier region.
The variable ∆ ln𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹𝐶𝐶 ,𝑡𝑡 is equal to zero if the region r is a frontier region in the given year.
This is because otherwise if r is a frontier region, there would be perfect collinearity between
the dependent variable and ∆ ln𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹𝐶𝐶 ,𝑡𝑡 and the estimate of 𝛽𝛽1 obtained would be biased
towards the share of frontier regions in our dataset (8%). By doing this, we obtain an estimate
for 𝛽𝛽1 that is estimated only over the non-frontier regions. We expect 𝛽𝛽1 to be positive as the
growth of the country frontier region has a positive effect on that of other regions in the same
country.
The productivity gap variable 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1 is defined as ln � 𝑃𝑃𝑟𝑟𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1𝑃𝑃𝑟𝑟𝑃𝑃𝑃𝑃𝐹𝐹𝐶𝐶,𝑡𝑡−1
�. The gap is equal to
zero at the frontier and becomes increasingly negative for regions farther away from the
frontier. 𝛽𝛽2 indicates the (hypothetical) effect of increasing the productivity gap by one unit
14
when policy indicators are set to zero. We expect the marginal effect of 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 to be
negative: as 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡 takes negative values, increasing 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡 is equivalent to
decreasing the distance to the frontier, which we expect to reduce the catching-up effect on
regional growth.
𝑃𝑃𝑃𝑃𝑃𝑃𝐶𝐶,𝑡𝑡−1 is the lagged level of the policy variable in country C. In order to facilitate the
interpretation of the coefficients, we normalise the structural policy indicators (see below).
Higher values of labour market legislation, state control and barriers to entrepreneurship or
barriers to trade variables are indicative of lower levels of regulation. Since we are interested
in the impact of nation-wide factors on different types of regions, and in particular on regions
depending on their distance to the country’s productivity frontier, each nation-wide policy
variable is interacted with the productivity gap variable. In our full specification, we include
all the structural policies together and the three macroeconomic indicators together. We obtain
the marginal effects of each policy, 𝛽𝛽3 + 𝛽𝛽4.𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1 as well as the corresponding
confidence intervals. 𝜁𝜁t are year-specific effects and γr region fixed-effects that account for
time-invariant regional factors that influence regional productivity growth.
We estimate the model using OLS with fixed effects. However the variable
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑟𝑟,𝑡𝑡−1includes a term which is the lagged dependent variable that in its transformed
form is correlated with the transformed error term. This would lead to estimation by fixed
effects being biased. In order to address this, and given the serial correlation in our variables
and the relatively small number of years in our dataset, we follow Blundell and Bond (1998)
and also estimate our model using a system-GMM method.
15
Noteworthy, the fact that we are using regional data enables to control for fixed-
effects without creating the usual collinearity problems with the country-wide policy
variables, which often do not display strong time variability.7
4. The data
Our data consist of a panel of 265 regions from 24 OECD countries8 defined at Territorial
Level 2 (TL2), taken from the OECD Regional Database and covering the period 1997 to
2007.9 We observe regional productivity since 1996, defined as GDP per worker, deflated
with base year 2000 and PPP adjusted in US dollars. We use this measure to compute yearly
regional productivity growth in percentages. Using the regional productivity data we are able
to identify the regions which are at the productivity frontier in their country in each year. This
allows us to compute the productivity growth of the frontier region (∆ ln 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹𝐶𝐶 ) and the
distance between a given region and the country frontier region (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃).
Turning to our country-year level policy and macroeconomic variables, our measures of
regulation are drawn from the OECD’s Product-Market Regulation (PMR) Database.10 The
7 This is a recurrent problem in many cross-country econometric growth studies using structural policy variables that do not display a time dimension or have low time variability. To circumvent this problem, researchers often have to revert to pooled or the GLS estimates that may generate bias on their own.
8 We do not use the full set of OECD countries and regions due to restrictions on data availability. The countries covered are Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Korea, Netherlands, Norway, Poland, Portugal, Spain, Sweden, Switzerland, UK and USA.
9 We prefer to restrain our sample to the pre-crisis period in order to capture the structural effects of policies. During an economic depression, structural reforms may have temporary deflationary effects counteracting their long-run benefits. This may generate a bias in the estimates of the gains from structural reform.
10 See Wolfl et al. (2009) for a detailed description of the PMR data as well as Koske et al. (2015) for a more recent update on the PMR data.
16
PMR indicators are a comprehensive and internationally comparable set of indicators that
measure the degree to which policies promote or inhibit competition in areas of the product
market where competition is viable. They measure the economy-wide regulatory and market
environments in 34 OECD countries as well as in Brazil, China, India, Indonesia, Russia and
South Africa. They are consistent across time and countries. In the original data, the values
for each indicator vary between 1 and 6, with higher values indicating higher levels of
restrictions. To make their interpretation more comparable with the other structural indicators,
we use the PMR with a reverse scale (i.e. higher values mean lower restrictions). Data are
gathered and the indicators calculated according to a common method, so as to ensure
consistency across time and comparability across space and across sectors. We use the three
main PMR components rather than the composite indicator because the latter combines quite
different dimensions, possibly affecting regional growth in a very diverse manner. The first is
State control that measures the extent of state ownership (scope of public enterprise, direct
control over business enterprises and government involvement in network sectors) and the
state’s involvement in business operations (price controls, use of command and control
regulation). The second are Barriers to entrepreneurship that measures regulatory and
administrative opacity, administrative burdens on start-ups and barriers to competition.
Finally, Barriers to trade and investment encompasses barriers to FDI, tariffs, discriminatory
procedures and regulatory barriers to trade and investment.
Our second structural policy area is that of labour market legislation. We use the
employment protection legislation (EPL) indicator constructed by the OECD’s Directorate for
Labour and Social Affairs, which varies from 1 to 6 (like for the PMR, we are using here a
reverse scale, which increases with labour market deregulation). EPL measures the procedures
and costs involved in dismissing or hiring workers, based on information provided by officials
17
in the OECD member countries and expert opinions from the International Labour
Organization (ILO). EPL indicators are available for all 35 OECD countries as well as for 39
non-OECD countries and territories. Two EPL indicators were available to us: the first one,
pertaining to regular contracts, was not retained in our analysis due to the lack of time and
cross-country variation in this measure; the second one, pertaining to temporary contracts,
presents considerable variation as this form of employment contracts has been subject to
reform in many countries in the period we consider.
In terms of macroeconomic variables, we consider trade openness, captured by total trade
flows as a percentage of GDP, government debt to GDP ratio, and inflation (measured as the
rise in the consumer price index). All macroeconomic data are drawn from OECD sources.
Table 1 provides summary statistics of the variables. Annual productivity growth has an
average value of 1.7% and ranges from -7.8% to +13.9%. Outlier observations with
productivity growth lower than -8% or greater than 14% have been removed from the sample,
although the frontier region growth is allowed to exceed these values. The productivity gap
takes on negative values. It is equal to zero for regions at the productivity frontier of their
country, and becomes increasingly negative for regions further away from the frontier.
5. Results
5.1 Impact of structural reforms
Table 2 reports our results on the effects of structural variables using regional fixed
effects, pooled OLS and system GMM methods. The first six columns in Table 2 report the
estimates from fixed effects regressions. In column (1), the positive and statistically
18
significant coefficient on Frontiergrowth indicates that increasing the productivity growth of
the country frontier region has a positive effect on the growth of the other regions, as
expected. The 0.05 coefficient means that a 1 percentage point increase in the annual
productivity growth of the country frontier region is associated with a 0.05 percentage point
increase in regional productivity growth. Although the effect is positive and significant, it is
of small magnitude: the frontier region is the region with the highest productivity level in the
country and may not have particularly high growth. We further investigate the pass-through
effect from the frontier in our robustness checks. Turning to the prodgap coefficient, this is
negative and statistically significant as expected and means that one standard deviation
increase in the distance to the frontier is associated with a 2.9% higher regional growth.
In column (6), we include the four policy variables together as well as all interaction terms
with the productivity gap in a fixed effects specification. The coefficient on Frontiergrowth
remains unchanged, while the coefficient of -13.95 on the productivity gap represents a
fictitious case where the value of all the structural variables would be at zero, meaning the
highest possible levels of regulation. The overall effect of the productivity gap however,
computed at average levels of the structural variables, is -9.2, meaning that increasing the
productivity gap by one standard deviation would increase regional growth by 4%.
Column (7) presents, as a means of comparison, the results obtained from a pooled OLS
estimation. Without region-specific effects, the effect of the frontier is no longer significant.
The cross-section effects dominate, capturing agglomeration forces rather than convergence
forces. Indeed, regional growth can be the result of convergence and agglomeration forces.
The former tend to favour lagging or poorer (often low-density, rural) areas. Conversely, the
agglomeration forces favour the dense, urbanised areas (see Garcilazo and Oliveira Martins,
19
2013). The fixed effects and system GMM with region effects methods however capture the
convergence forces.
Figure 1 below represents the marginal effects with 95% confidence intervals of each of
our four structural reforms, based on the estimations underlying the results reported in column
(6) of Table 2. Each panel of Figure 1 represents, along the left vertical axis, the marginal
effect of a one-point increase in the value of the variable (i.e. greater deregulation) on regional
productivity growth, at different levels of the productivity gap with the frontier region. In
addition, each panel shows, according to the right vertical axis, the histogram of the
productivity gap variable. This enables us to see the magnitude and the significance of the
marginal effects, over a meaningful range of the productivity gap.
Figure 1 shows that deregulating temporary contracts EPL has a small negative effect on
the productivity growth of regions that are near the frontier, whilst the effect is large and
positive for a relatively small number of regions with a productivity gap lower than -0.6. The
positive effect on growth increases in the productivity gap. For the middle range of regions,
the marginal effect of deregulating EPL is not significantly different from zero. Deregulating
State Control seems to have no significant effect on regions close to the frontier, while there
is an increasingly negative impact on growth points for regions with a productivity gap
smaller than -0.4 as they get further from the frontier. Turning to barriers to entrepreneurship,
in this specification deregulation appears to have no significant effect on the growth of most
regions (with a productivity gap greater than -0.6) while there is a positive and significant
effect on lagging regions that increases with distance to the frontier. Finally, reducing barriers
to trade and investment has a negative and significant effect for the middle range of regions,
with a larger negative impact on regions as their distance to the frontier increases.
20
Figure 1: Marginal effects of structural reforms, OLS
As explained in Section 3, our preferred estimation method is system GMM. The
corresponding results in column (8) of Table 2 confirm the pass-through effect of the frontier
region with a highly significant coefficient of +0.07 on the frontier region growth. The graphs
in Figure 2 below represent the marginal effects obtained from this specification. The system
GMM estimation yields very similar results for EPL. Intuitively, this differential impact of
labour market legislation among different kinds of regions can be driven by agglomeration
effects. Labour markets in regions close to the frontier are likely to be “thicker”, with larger
and more diverse populations of workers. Other things being equal, labour-market rigidities
are likely to be less costly in thicker labour markets and in those that are better endowed with
skills, because skill supply and matching are likely to be easier under any given regulatory
21
regime. Regulatory rigidities in labour markets are likely to exact a much higher price in
regions farther from the frontier.
Figure 2: Marginal effects of structural reforms, GMM
In this preferred specification, the effects of deregulating state control, barriers to
entrepreneurship and barriers to trade and investment on regional growth are no longer
significant over the entire range of productivity gap. We now investigate this further by using
an alternative definition of the frontier as well as checking for sample selection issues.
Robustness checks
Our results may be sensitive to specificities of our data or to the way in which we have
defined a single frontier region for each country. We first consider the possibility that the
22
presence of frontier regions in our dataset for analysis may influence our results. We remove
from our data all the observations that correspond to a country’s frontier region in any given
year. As a result, we are left with 2,281 observations for which the productivity gap is not
zero. Results are presented in Table 3. The marginal effects computed from the results of the
OLS and system GMM specifications of columns (6) and (7) respectively are shown
graphically in Figures 3 and 4 below.
23
Figure 3: Exclusion of frontier regions, OLS
Figure 4: Exclusion of frontier regions, system GMM
24
Results from the OLS with fixed effects specification are similar to those obtained
previously. Focusing on our GMM results (Figure 4), the positive effect of labour market
deregulation on lagging regions is confirmed, while deregulating barriers to trade and
investment now has a positive and significant effect on lagging regions, that increases with
distance to the frontier.
Similarly, in Table 4 we remove the observations corresponding to the three regions with
the highest productivity in each country in each year. This estimates the effects of the frontier
and of structural policies on the periphery. The pass-through and catching-up effects are
robust as indicated by the coefficients in column (1). Figures 5 and 6 below represent the
marginal effects obtained from the OLS and GMM estimations of columns (6) and (7).
Again, the fixed effects results are very similar to those obtained from the whole sample,
with more pronounced effects of EPL and barriers to trade (Figure 5). Turning to the GMM
results in Figure 6, we find that again reducing labour market rigidities on temporary
contracts has a positive effect on growth for lagging regions; reducing barriers to
entrepreneurship has a positive and significant effect for a range of regions close to the
productivity frontier, with the largest effect closer to the frontier, and reducing barriers to
trade and investment still has a clear positive and significant impact on growth for lagging
regions. The effect of state control again is not significant for lagging regions although there
is a negative and significant effect of deregulating state control on growth near the frontier.
25
Figure 5: Excluding the top three regions, OLS.
Figure 6: Excluding the top three regions, GMM.
26
Another possible important bias may come from our definition of the frontier regions.
Instead of using the region with the highest productivity in region r’s country, we propose an
alternative definition of the frontier by creating a synthetic frontier where the productivity is
equal to the average productivity in the three most productive regions in the country in a
particular year. We also compute an alternative measure of the distance to the frontier, which
is now equal to the difference between the log of productivity of region r and the log of the
average productivity of the three most productive regions. We then remove from our dataset
the three regions that were used in order to create the synthetic frontier region. This is to
avoid having a positive productivity gap between the highest productivity region and the
synthetic average and to capture the effect of this synthetic frontier and of the policies on
those regions that are not part of this frontier.
The results in Table 5 show that regional growth increases in the growth of the synthetic
frontier, it also increases in the distance to the synthetic frontier, as presented in our base
results in Table 2. It is noteworthy that the pass-through effect estimated here is much larger
than that identified in Table 2 with the full sample or even in Table 4 without the top three
regions: a one percentage point increase in the growth of the synthetic frontier is associated
with a 0.3 point increase in regional growth for the other regions. For all regions except the
three most productive, the pass-through of the economic core of the country is more
meaningful than that of the single frontier region. We therefore focus on this particular
definition of the frontier and the results associated with it.
The marginal effects are depicted in Figure 7 (OLS with region fixed effects) and Figure 8
(GMM) below. Given the new definition of the frontier and the exclusion of the top three
regions, the distribution of the productivity gap variable has changed, with a higher minimum
27
(-2.94) and a lower maximum (-0.003). This is reflected in the horizontal axis and in the
histograms represented on both figures.
Figure 7: Alternative frontier region, excluding the top three regions, OLS.
28
Figure 8: Alternative frontier region, excluding the top three regions, GMM.
Figure 8 above confirms the patterns found on EPL. For lagging regions, the reduction in
rigidities in temporary contracts acts in favour of reallocation and labour market
experimentation, while for regions near the frontier, deregulating temporary contracts might
in fact prevent the accumulation of specific capital in firms. For a region with a productivity
gap of -0.7, a one-point increase in the EPL indicator (just less than one standard deviation) is
associated with an increase in productivity growth of 1 percentage point. For a region with a
productivity gap of -1.5, the same amount of deregulation is associated with an increase in
growth of 4.8 points.
While the marginal effect of deregulating state control is positive and increasing in
distance to the frontier, it is still insignificant. This could be due to the different aspects of
state control in this variable having counteracting effects. We also find that reducing barriers
29
to entrepreneurship has a positive and significant effect on regional growth for a large range
of regions with productivity gap greater than -0.9. This effect increases for regions closer to
the frontier. For example, for a region with a productivity gap of -0.5, increasing the policy
indicator by one point (roughly two standard deviations) is associated with a 3.4 point
increase in productivity growth, whilst for a region with a productivity gap of -0.3, the
marginal effect is 4.3 growth points. For regions farther from the frontier, contrary to our
expectations, the effect of reducing barriers to entrepreneurship is insignificant: The
discouragement effect may be prevailing in these regions as competition increases with the
lowering of these barriers.
Turning to barriers to trade and investment, as expected, a reduction of this type of
rigidities has a positive and significant effect on growth for the range of regions farther away
from the frontier (with a productivity gap less than -0.5). For example, for a region with a
productivity gap of -0.5, increasing the policy indicator by one point (here also roughly two
standard deviations) is associated with a 1.2 point increase in productivity growth. For a
region with a productivity gap of -1, the effect of the same increase in the policy indicator is a
4.64 point increase in productivity growth. Although the effects estimated on lagging regions
are very large, they remain of a more realistic magnitude over the main range of regions in
our dataset.
Finally, given the time period of our analysis it could be argued that Poland, Ireland and
Spain could have an overly important weight on the results. During the pre-crisis period, these
three countries experienced particularly rapid growth and undertook important structural
reforms. Accordingly, we replicated our base estimates after removing Polish, Irish and
Spanish regions from our sample (see Table 6). Marginal coefficients represented in Figures
9-11 show that our results are robust to the exclusion of these countries.
30
Figure 9: Excluding Spanish regions.
−20
24
68
1012
1416
Mar
gina
l effe
ct o
f EPL
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−12
−10
−8−6
−4−2
02
46
810
12
Mar
gina
l effe
ct o
f Sta
te C
ontr
ol
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−6−5
−4−3
−2−1
01
23
Mar
gina
l effe
ct o
f Bar
riers
to E
ntre
pren
eurs
hip
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−4−2
02
46
810
1214
16
Mar
gina
l effe
ct o
f Bar
riers
to T
rade
and
Inve
stm
ent
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
Structural policies − GMM − Spanish regions excludedMarginal effects on productivity growth
31
Figure 10: Excluding Polish regions.
Figure 11: Excluding Irish regions.
−20
24
68
1012
1416
Mar
gina
l effe
ct o
f EPL
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−12
−10
−8−6
−4−2
02
46
8
Mar
gina
l effe
ct o
f Sta
te C
ontr
ol
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−12
−10
−8−6
−4−2
02
4
Mar
gina
l effe
ct o
f Bar
riers
to E
ntre
pren
eurs
hip
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−50
510
1520
2530
3540
Mar
gina
l effe
ct o
f Bar
riers
to T
rade
and
Inve
stm
ent
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
Structural policies − GMM − Polish regions excludedMarginal effects on productivity growth
−20
24
68
1012
1416
Mar
gina
l effe
ct o
f EPL
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−10
−8−6
−4−2
02
46
810
Mar
gina
l effe
ct o
f Sta
te C
ontr
ol
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−6−5
−4−3
−2−1
01
23
Mar
gina
l effe
ct o
f Bar
riers
to E
ntre
pren
eurs
hip
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
−4−2
02
46
810
1214
16
Mar
gina
l effe
ct o
f Bar
riers
to T
rade
and
Inve
stm
ent
02
46
810
1214
1618
20Pe
rcen
tage
of o
bser
vatio
ns
−4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0
Productivity gap
Structural policies − GMM − Polish regions excludedMarginal effects on productivity growth
32
5.3 Impact of macroeconomic factors
We now turn to the analysis of the macroeconomic variables. Table 7 reports our base results
using the full dataset and the single frontier region. The first three robustness checks,
including the alternative definition of the frontier, are offered in Table 8. Table 9 shows the
results obtained once we exclude Polish, Spanish and Irish regions. Again, the coefficient on
frontier region growth is highly significant and robust across specifications. We show below
in Figures 12-18 the marginal coefficients obtained from our preferred GMM estimations.
Figures from the OLS with region fixed effects estimations are included in the appendix.
Figure 12 corresponds to the results from column (7) in Table 7. As expected,
increasing trade exposure has a positive and significant effect on regions with large
productivity gaps (values less than -1.3). For a region with a productivity gap of -1.5,
increasing trade openness by one standard deviation is associated with a 4.32 point increase in
growth. However, we also find a negative and significant effect on regions close to the
frontier (values greater than -0.6). For a region with a productivity gap of -0.5, increasing
trade openness by one standard deviation is associated with a 1.76 point decrease in growth.
These results are similar to and consistent with those found on the effects of liberalising trade
and investment (Figures 2 and 8) and they are robust to our modifications in the sample and
the definition of the frontier presented in Figures 13-18 and Tables 8 and 9. They however go
against the theory that, in regions close to the productivity frontier, heightened competition
brought about by greater trade and investment would stimulate the escape entry effect, raising
productivity. In our less stringent, alternative definition of the frontier however (Figure 15),
the negative effect of trade exposure is only significant for a smaller range of regions very
close to the frontier. Overall, both estimates on structural policies and on macroeconomic
33
factors support the hypothesis that exposure to international trade has an enhancing effect on
regional productivity growth, at least for regions far enough from the frontier. As explained in
Section 2, a possible explanation could be related to the role of the tradable sector as an
engine of unconditional growth (Rodrik, 2013).
Figure 12: Marginal effects of macroeconomic variables, GMM.
Government debt has a positive and significant effect on regional growth that
increases with distance to the frontier.11 In the base results, for a region with a productivity
gap of -0.5, increasing government debt (as a percentage of GDP) by one standard deviation
is associated with a 1.1 point increase in growth. For a region farther from the frontier, with a
productivity gap of -1.5, the same increase in government debt would yield a 1.7 point
increase in growth. In our alternative specifications, although the effect of debt remains
11 Note that this is the effect of increases in debt over time due to the inclusion of region fixed effects.
34
positive and significant and of similar magnitude for a range of regions not too far from the
frontier, the effect is now (very mildly) decreasing with distance to the frontier. As shown in
Figure 15, when we use the average frontier and exclude the top three regions, the effect of
government debt is positive and significant for regions with values of the productivity gap
greater than -1.3 and the effect is no longer precisely estimated for the small number of
regions farther from the frontier. For a region with a productivity gap of -0.5, increasing
government debt by one standard deviation is associated with a 1.5 point increase in growth.
As can be seen from our results in Figures 12-18, the positive effect of government debt on
regional growth for a meaningful range of regions is robust across our specifications. It is
puzzling that, when we remove the frontier regions, we are not able to find a significant
positive effect of debt on lagging regions, whilst these regions are most likely to be on the
receiving end of public investment.
Finally, throughout our estimations, we do not find evidence of a significant role of
inflation in regional growth. This is not due to a lack of time variation in our data. Rather, as
explained in Section 2, this variable is included as one of the macroeconomic imbalance
indicators but there is no clear prediction on its differential effect on regional growth. For
example, inflation in lagging regions may affect export competitiveness due to the price of
their exports increasing, while at the same time reducing the cost of imported intermediates. If
lagging regions are particularly sensitive to increases in trade openness, as our results show,
we may expect inflation increases to affect their productivity growth, but the net effect would
be unclear.
35
Figure 13: Excluding frontier regions.
Figure 14: Excluding top three regions.
36
Figure 15: Alternative frontier region.
Figure 16: Excluding Spanish regions.
37
Figure 17: Excluding Polish regions.
Figure 18: Excluding Irish regions.
38
6. Conclusions
Our analysis measures the effects of country-wide macroeconomic and structural
factors on regional performance. We consider both the effects of regulatory policies and
macroeconomic factors on regional productivity using a panel covering 265 regions from 24
OECD countries over the period 1997-2007 representing roughly three business cycles. We
find strong statistical links between economy-wide macroeconomic and structural policies and
regional productivity that are not homogenous across regions; they tend to vary with respect
to the distance of regions to their national productivity frontier, with typically lagging regions
being the most affected by rigidities in product and labour markets. In addition, our analysis
finds evidence of a catching up effect with faster productivity growth in lagging regions and
evidence of a pass-through effect where growth in the frontier regions boosts the productivity
growth of the other regions. In particular, the effect of the frontier is more pronounced for
peripheral regions than for regions immediately following the frontier in terms of
productivity. We also find that the top three most productive regions, which we can consider
the core of each country, have a stronger effect on regional growth than the single most
productive region.
Our estimates reveal that deregulating employment protection legislation in temporary
contracts or reducing the level of barriers to trade and investment has a strong positive effect
on the growth of lagging regions and that this increases in distance to the frontier. The
increase in regional growth from deregulating EPL of temporary contracts by one point on the
indicator scale (0.8 of a standard deviation) varies from 1 to 4.8 growth points between a
region at a productivity gap of -0.7 and another region at a productivity gap of -1.5.
Deregulating barriers to trade and investment by one point (roughly two standard deviations)
39
is associated with 1.2 point higher regional growth for a region with a productivity gap of -0.5
and 4.64 points for a region with a productivity gap of -1.
The effect of international trade is confirmed in our separate results on macroeconomic
factors: increasing trade openness (expressed as trade volume as a percentage of GDP) at the
country level increases regional growth of regions that are far from the productivity frontier,
with a more pronounced effect for lagging regions.
On the other hand, reducing barriers to entrepreneurship increases productivity growth for
regions closer to the frontier: for a region with a productivity gap of -0.5, deregulating
barriers to entrepreneurship by one point (roughly two standard deviations) is associated with
a 3.4 point increase in growth. This effect is larger for regions nearer the frontier but in our
GMM results we do not find a significant effect on regions that are the most lagging. We also
fail to find a significant effect of reducing state control on regional growth, across the range
of regions with our GMM specification.
Our findings also indicate that government debt has a positive and significant effect on
growth for most regions with values of the productivity gap greater than -1.3. For a region
with a productivity gap of -0.5, increasing government debt by one standard deviation is
associated with a 1.5 point increase in growth and the magnitude of the effect is larger for
regions nearer the frontier.
These findings reveal a strong link between the national and the regional dimension which
carries important policy implications. First, they help to understand how national factors have
a differentiated impact across regions enabling us to better assess their overall effects. Our
results also suggest that structural and macroeconomic policies should account for these
40
regional effects in their design by complementing these policies with policies targeted to
specific regions to enhance their effects or restrain their negative effects.
Finally, arguments against regulatory reform have been made on the basis of harming
vulnerable or strategic regions. Our results do not warrant these views. On the contrary,
macro-structural policies tend to support catching-up of the lagging regions. This also
provides some basis to justify conditionalities associated with structural reforms. Our results
tell us that regulatory effects tend to vary according to regions’ distance to the frontier, they
also highlight the fact that different forms of regulation have different regional impacts.
41
References
Acemoglu, D., Aghion, P. and F. Zilibotti (2006), ‘‘Distance to Frontier, Selection and Economic Growth,’’ Journal of the European Economic Association 4(1): 37–74.
Aghion, P., Blundell, R., Griffith, R., Howitt, P. and S. Prantl (2004), ‘‘Entry and Productivity Growth: Evidence from Microlevel Panel Data,’’ Journal of the European Economic Association 2: 265–276.
Aghion, P., Burgess, R., Redding, S. and F. Zilibotti (2006), ‘‘The Unequal Effects of Liberalization: Evidence from Dismantling the License Raj in India,’’ American Economic Review 98: 1397–1412.
Aghion, P., Harris, C., Vickers, J. (1997), “Competition and growth with step-by-step innovation: An example”. European Economic Review, Papers and Proceedings 41: 771-782.
Aghion, P., Howitt, P. (1998), “Endogenous Growth Theory”. MIT Press, Cambridge, MA.
Aghion, P. and P. Howitt (2006), ‘‘Joseph Schumpeter Lecture: Appropriate Policy Growth: A Unifying Framework,’’ Journal of the European Economic Association 4: 269–314.
Ahluwalia, M. (2000), “State level performance under economic reforms in India”, Centre for Research on Economic Development and Policy Reform Conference on Indian Economic Prospects: Advancing Policy Reform.
Barro, R., Sala-i-Martin, X. (2004), “Economic Growth” (Second Edition). MIT Press, Cambridge, MA.
Bassanini, A., Scarpetta, S. (2001), “The driving forces of economic growth: panel data evidence for the OECD countries”. OECD Economic Studies No. 33.
Becker, S., Egger. P., von Ehrlich, M. (2012), “Too much of a good thing? On the growth effects of the EU’s regional policy”. European Economic Review 56(4): 648-668.
Becker, S., Egger. P., von Ehrlich, M. (2013), “Absorptive capacity and the growth and investment effects of regional transfers: A regression discontinuity design with heterogeneous treatment effects”. American Economic Journal: Economic Policy 5(4): 29-77.
Boldrin, M., Canova, F. (2001), “Inequality and convergence in Europe’s regions: reconsidering European regional policies”. Economic Policy, April 2001: 207-253.
42
Bourlès, R., Cette, G., Lopez, J., Mairesse, J., Nicoletti, G. (2013), “Do product market regulations in upstream sectors curb productivity growth? Panel data evidence for OECD countries”. The Review of Economics and Statistics 95(5): 1750-1768.
Breidenbach, P., Mitze, T., Schmidt, C. (2016), “EU Structural Funds and Regional Income Convergence - a Sobering Experience”. CEPR Discussion Paper Series no. 11210.
Breinlich, H., Ottaviano, G., Temple, J. (2014), “Regional Growth and Regional Decline”. In: Aghion, P., Durlauf, S. (eds), Handbook of Economic Growth, vol. 2, Elsevier.
Chandra, A., Thompson, E. (2000), “Does public infrastructure affect economic activity? Evidence from the rural interstate highway system”. Regional Science and Urban Economics 30(4): 457-490.
Che, N. X., Spilimbergo, A. (2012), “Structural reforms and regional convergence”. CEPR Discussion Paper No. 8951.
Crescenzi, R., Rodríguez-Pose, A. (2012), "Infrastructure and regional growth in the European Union". Papers in Regional Science 91(3): 487-513.
Desmet, K., Rossi-Hansberg, E. (2010), “On Spatial Dynamics”. Journal of Regional Science 50(1): 43-63.
Duranton, G., Turner, M. (2011), “The fundamental law of road congestion: Evidence from US cities”. American Economic Review 101(6): 2616-52.
Garcilazo, E., Oliveira Martins, J. (2013), “The Contribution of Regions to Aggregate Growth in the OECD”. Economic Geography 91(2): 205–221.
Glaeser, E. L., Scheinkman, J. A., Shleifer, A. (1995), “Economic growth in a cross-section of cities”. Journal of Monetary Economics 36(1): 117-143.
Griffith, R., Redding, S. and J. van Reenen (2004), ‘‘Mapping the Two Faces of R&D: Productivity Growth in a Panel of OECD Countries,’’ Review of Economics and Statistics 86(4): 883–895.
Henderson, V., Kuncoro, A., Turner, M. (1995), “Industrial development in cities”. Journal of Political Economy 103: 1067-1090.
Kim, S. T. (1997), “The role of local public sectors in regional economic growth in Korea”. Asian Economic Journal 11(2): 155-168.
Koske, I., Wanner, I., Bitetti, R., Barbiero, O. (2015), "The 2013 update of the OECD's database on product market regulation: Policy insights for OECD and non-OECD
43
countries". OECD Economics Department Working Papers, No. 1200, OECD Publishing, Paris.
Lucas, R. E. (1988), “On the mechanics of economic development”. Journal of Monetary Economics 22: 3-42.
Martin, P., Ottaviano, G. I. P. (1999), "Growing locations: Industry location in a model of endogenous growth". European Economic Review 43(2): 281-302.
McCann, P., Ortega-Argiles, R. (2013), “Transforming European regional policy: a results-driven agenda and smart specialization”. Oxford Review of Economic Policy 29(2): 405–431.
Michaels, G. (2008), “The effect of trade on the demand for skill – Evidence from the interstate highway system”. Review of Economics and Statistics 90(4): 683-701.
Minerva, G. A., Ottaviano, G. I. P. (2009), “Endogenous growth theories: Agglomeration benefits and transportation costs”. In: Capello, R., Nijkamp, P. (eds), Handbook of Regional Growth and Development Theories, Edward Elgar: Cheltenham.
Minniti, A., Parello, C. (2011), “Trade integration and regional disparity in a model of scale-invariant growth”. Regional Science and Urban Economics 41: 20-31.
Nicoletti, G., Scarpetta, S. (2003), “Regulation, productivity and growth. OECD Evidence”. Economic Policy April 2003: 9-72.
OECD (2009), How Regions Grow: Trends and Analysis. OECD, Paris.
OECD (2011), Regional Outlook, OECD Publishing, Paris.
OECD (2015), The Future of Productivity, OECD Publishing, Paris.
OECD (2016), Regional Outlook, OECD Publishing, Paris.
Rauch, J. E. (1993), “Productivity gains from geographical concentration of human capital: evidence from the cities”. Journal of Urban Economics 34: 380-400.
Roberts, M. (2010), "Endogenous Regional Growth: A Critical Survey". In: Setterfield, M. (ed), Handbook of Alternative Theories of Economic Growth, Edward Elgar: Cheltenham.
Rodrik, D. (2013), Unconditional Convergence in Manufacturing. The Quarterly Journal of Economics: 65–204.
Rodriguez-Pose, A., Fratesi, U. (2003), “Between development and social policies: the impact of European structural funds in Objective 1 regions”. EEG Working Paper no. 28/2003.
44
Romer, P. (1986), “Increasing returns and long-run growth”. Journal of Political Economy 94: 1002-1037.
Romer, P. (1990), “Endogenous technological change”. Journal of Political Economy 98: 71-101.
Sianesi, B., Van Reenen, J. (2003), “The returns to education: Macroeconomics”. Journal of Economic Surveys 17(2):157-200.
Solow, R. M. (1956), “A contribution to the theory of economic growth”. Quarterly Journal of Economics 70: 65-94.
Sachs, J., Bajpai, N., Ramiah, A. (2002), “Understanding regional economic growth in India”. CID Working Paper No.88.
Sala-i-Martin, X. (1996), "Regional cohesion: Evidence and theories of regional growth and convergence". European Economic Review 40(6): 1325-1352.
Stephens, H., Partridge, M., Faggian, A. (2013), “Innovation, entrepreneurship and economic growth in lagging regions”. Journal of Regional Science 53(5): 778-812.
Swan, T. (1956), “Economic growth and capital accumulation”. Economic Record 32: 334-61.
Wolfl, A., Wanner, I., Kozluk, T., Nicoletti, G. (2009), “Ten years of product market reform in OECD countries – Insights from a revised PMR indicator”. OECD Economics Department working paper No. 695.
45
Tables
Table 1: Summary statistics
Mean Std. dev. Min Max
Productivity levels (US$ constant prices) 60730.7 36522.7 14169.69 609012.8
Productivity growth 1.69 3.04 -7.84 13.9
Frontier region growth 3.66 6.42 -10.57 64.68
Productivity gap -0.43 0.43 -3.7 0
EPL temporary contracts 4.47 1.21 1.25 5.75
PMR state control 3.52 0.79 1.76 4.85
PMR barriers to entrepreneurship 3.71 0.54 2.55 4.56
PMR barriers to trade and investment 5.14 0.53 2.85 5.77
Trade openness (X+M)/GDP 31.17 16.05 9.48 92.37
Debt as % of GDP 66.77 28.13 12.6 175.27
Inflation rate (in %) 2.36 1.65 -0.9 14.15
Source: Own calculations using data from OECD Regional Database, OECD PMR Database, OECD EPL Database and OECD Economic Outlook Database.
46
Table 2: Structural reforms and regional growth
Dependent variable: regional
productivity growth (1) (2) (3) (4) (5) (6) (7) (8)
FE FE FE FE FE FE Pooled OLS
System
GMM
Frontiergrowth 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05** 0.03 0.07***
(0.018) (0.018) (0.018) (0.019) (0.019) (0.019) (0.017) (0.021)
Productivity gap t-1 -6.80*** 1.48 -2.37 -4.77*** -5.52* -13.95** 7.28** 12.59
(1.310) (1.877) (2.499) (1.419) (2.952) (5.996) (2.889) (9.414)
EPL temp t-1
-0.95***
-1.19*** 0.30* -1.51***
(0.254)
(0.292) (0.169) (0.403)
EPL temp X prodgap t-1
-2.27***
-2.51*** -0.66* -3.95***
(0.384)
(0.459) (0.349) (0.607)
Statecontrol t-1
-0.72**
0.20 -0.22 -0.45
(0.339)
(0.463) (0.220) (0.643)
Statecontrol X prodgap t-1
-1.29**
2.87*** 0.58 -0.71
(0.609)
(0.902) (0.440) (1.561)
Barrierstoentr t-1
-0.43
0.18 -0.75*** 1.02
(0.408)
(0.516) (0.259) (0.716)
Barrierstoentr X prodgap t-1
-0.64***
-1.08** -0.15 0.91
(0.163)
(0.456) (0.265) (0.667)
Barrierstotrade t-1
-0.19 0.17 -0.94*** -1.30**
(0.355) (0.631) (0.263) (0.640)
Barrierstotrade X prodgap t-1
-0.25 1.92* -1.18** -2.37*
(0.517) (1.161) (0.576) (1.361)
N 2,492 2,492 2,492 2,492 2,492 2,492 2,492 2,492
R2 0.087 0.103 0.089 0.091 0.087 0.110 0.112
AR(1)
0.000
AR(2)
0.115
Sargan
0.000
Number of instruments
337
Number of regions 265 265 265 265 265 265 265
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
47
Table 3: Sensitivity of the results to the exclusion of frontier regions
Dependent variable: regional
productivity growth (1) (2) (3) (4) (5) (6) (7)
FE FE FE FE FE FE System GMM
Frontiergrowth 0.06*** 0.06*** 0.06*** 0.06*** 0.06*** 0.06*** 0.08***
(0.019) (0.019) (0.019) (0.020) (0.019) (0.019) (0.022)
Productivity gap t-1 -7.31*** 1.68 -1.83 -5.37*** -6.25** -16.90*** 21.74**
(1.402) (1.948) (2.635) (1.511) (3.089) (6.436) (10.801)
EPL temp t-1
-1.12***
-1.48*** -1.95***
(0.259)
(0.337) (0.439)
EPL temp X prodgap t-1
-2.48***
-2.85*** -4.48***
(0.390)
(0.511) (0.678)
Statecontrol t-1
-0.85**
0.44 -0.72
(0.340)
(0.502) (0.696)
Statecontrol X prodgap t-1
-1.62**
3.14*** -1.43
(0.639)
(0.969) (1.702)
Barrierstoentr t-1
-0.34
0.52 1.70**
(0.439)
(0.592) (0.838)
Barrierstoentr X prodgap t-1
-0.65***
-1.15** 1.61**
(0.172)
(0.500) (0.774)
Barrierstotrade t-1
-0.12 0.52 -2.06**
(0.376) (0.703) (0.865)
Barrierstotrade X prodgap t-1
-0.21 2.47** -3.89**
(0.543) (1.245) (1.698)
N 2,281 2,281 2,281 2,281 2,281 2,281 2,281
R2 0.094 0.114 0.098 0.099 0.095 0.123
AR(1)
0.000
AR(2)
0.182
Sargan
0.000
Number of instruments
321
Number of regions 249 249 249 249 249 249 249
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
48
Table 4: Sensitivity of the results to the exclusion of the top 3 productivity regions
Dependent variable:
regional productivity
growth (1) (2) (3) (4) (5) (6) (7)
FE FE FE FE FE FE
System
GMM
Frontiergrowth 0.06*** 0.06*** 0.06*** 0.06*** 0.06*** 0.07*** 0.09***
(0.020) (0.019) (0.020) (0.020) (0.020) (0.020) (0.023)
Productivity gap t-1 -6.37*** 2.90 1.25 -5.02*** -5.69 -16.23** 43.15***
(1.603) (2.035) (2.829) (1.706) (3.453) (7.987) (16.085)
EPL temp t-1
-1.38***
-1.92*** -2.48***
(0.275)
(0.388) (0.497)
EPL temp X prodgap t-1
-2.62***
-3.12*** -4.69***
(0.397)
(0.551) (0.749)
Statecontrol t-1
-1.25***
-0.11 -2.50***
(0.323)
(0.543) (0.823)
Statecontrol X prodgap t-1
-2.28***
2.57** -4.64*
(0.690)
(1.140) (2.458)
Barrierstoentr t-1
0.04
1.64** 3.72***
(0.479)
(0.725) (0.992)
Barrierstoentr X prodgap t-1
-0.54***
-0.79 3.41***
(0.167)
(0.611) (1.087)
Barrierstotrade t-1
-0.03 0.69 -3.62***
(0.436) (0.936) (1.265)
Barrierstotrade X prodgap t-1
-0.13 2.77* -6.75***
(0.607) (1.577) (2.526)
N 1,856 1,856 1,856 1,856 1,856 1,856 1,856
R2 0.098 0.124 0.105 0.104 0.098 0.139
AR(1)
0.000
AR(2)
0.123
Sargan
0.000
Number of instruments
288
Number of regions 216 216 216 216 216 216 216
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors
are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
49
Table 5: Sensitivity of the results to an alternative definition of frontier regions
Dependent variable:
regional productivity growth (1) (2) (3) (4) (5) (6) (7)
FE FE FE FE FE FE System GMM
Frontiergrowth 0.27*** 0.26*** 0.27*** 0.30*** 0.28*** 0.30*** 0.32***
(0.035) (0.035) (0.036) (0.035) (0.036) (0.034) (0.038)
Productivity gap t-1 -16.71*** -8.65*** -10.81*** -17.76*** -14.88*** -21.50** 30.86
(1.954) (2.360) (4.046) (1.976) (4.065) (8.996) (20.563)
EPL temp t-1
-1.07***
-1.67*** -2.39***
(0.275)
(0.374) (0.497)
EPL temp X prodgap t-1
-2.09***
-2.78*** -4.81***
(0.487)
(0.663) (0.811)
Statecontrol t-1
-0.95***
-0.85 -2.64***
(0.356)
(0.556) (0.901)
Statecontrol X prodgap t-1
-1.79*
1.71 -7.08*
(0.984)
(1.753) (3.688)
Barrierstoentr t-1
1.63***
3.45*** 5.62***
(0.435)
(0.609) (0.807)
Barrierstoentr X prodgap t-1
-0.13
0.37 4.46***
(0.212)
(0.639) (1.309)
Barrierstotrade t-1
0.15 0.23 -2.25**
(0.411) (0.708) (0.964)
Barrierstotrade X prodgap t-1
-0.41 1.27 -6.89***
(0.670) (1.461) (2.613)
N 1,856 1,856 1,856 1,856 1,856 1,856 1,856
R2 0.184 0.194 0.187 0.195 0.185 0.219
AR(1)
0.000
AR(2)
0.477
Sargan
0.000
Number of instruments
288
Number of regions 216 216 216 216 216 216 216
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
50
Table 6: Sensitivity of the results to the removal of Spanish, Irish and Polish regions
Dependent variable: regional
productivity growth Without Spain Without Poland Without Ireland
(1) (2) (3) (4) (5) (6)
FE
System
GMM FE
System
GMM FE
System
GMM
Frontiergrowth 0.04** 0.06*** 0.04** 0.06*** 0.05** 0.07***
(0.020) (0.023) (0.019) (0.022) (0.019) (0.021)
Productivity gap t-1 -14.86** 11.53 -24.73** 22.11 -14.11** 11.77
(5.747) (9.411) (10.877) (16.490) (5.981) (9.382)
EPL temp t-1 -1.45*** -1.64*** -1.44*** -1.87*** -1.23*** -1.52***
(0.300) (0.419) (0.293) (0.370) (0.293) (0.405)
EPL temp X prodgap t-1 -2.75*** -4.01*** -2.12*** -3.80*** -2.54*** -3.96***
(0.468) (0.628) (0.473) (0.589) (0.461) (0.609)
Statecontrol t-1 1.30* 0.29 -0.27 -0.51 0.23 -0.41
(0.675) (0.939) (0.426) (0.586) (0.465) (0.643)
Statecontrol X prodgap t-1 3.96*** -0.08 2.33*** 0.15 2.99*** -0.56
(1.007) (1.711) (0.855) (1.386) (0.905) (1.562)
Barrierstoentr t-1 -0.06 0.83 0.26 1.45* 0.17 1.01
(0.518) (0.732) (0.513) (0.746) (0.517) (0.716)
Barrierstoentr X prodgap t-1 -1.20*** 0.77 -1.66** 1.92 -1.09** 0.86
(0.436) (0.668) (0.816) (1.193) (0.454) (0.664)
Barrierstotrade t-1 0.20 -1.16* 2.39*** -0.84 0.13 -1.28**
(0.583) (0.644) (0.828) (1.242) (0.629) (0.637)
Barrierstotrade X prodgap t-1 1.78 -2.29* 4.57* -5.33 1.91* -2.28*
(1.081) (1.317) (2.360) (3.363) (1.156) (1.355)
N 2,309 2,309 2,355 2,355 2,474 2,474
R2 0.099
0.115
0.109
AR(1)
0.000
0.000
0.000
AR(2)
0.205
0.148
0.116
Sargan
0.000
0.000
0.000
Number of instruments
318
321
335
Number of regions 246 246 249 249 263 263
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
51
Table 7: Macroeconomic factors and regional growth
Dependent variable: regional
productivity growth (1) (2) (3) (4) (5) (6) (7)
FE FE FE FE FE Pooled OLS
System
GMM
Frontiergrowth 0.05*** 0.05*** 0.06*** 0.05*** 0.05*** 0.03** 0.06***
(0.018) (0.018) (0.019) (0.019) (0.019) (0.017) (0.020)
Productivity gap t-1 -6.80*** -0.45 -5.91*** -7.74*** -2.76 0.97 2.48
(1.310) (2.240) (1.375) (1.348) (1.996) (0.825) (3.598)
Trade t-1
-0.19***
-0.18*** 0.01 -0.30***
(0.048)
(0.045) (0.008) (0.073)
Trade X prodgap t-1
-0.19***
-0.12** -0.03* -0.38***
(0.058)
(0.048) (0.016) (0.109)
Debt t-1
0.01
0.02* -0.01*** 0.03*
(0.012)
(0.012) (0.003) (0.016)
Debt X prodgap t-1
-0.03***
-0.02* -0.01* -0.02
(0.006)
(0.008) (0.005) (0.013)
Inflation t-1
0.09 -0.04 0.11 -0.00
(0.072) (0.080) (0.074) (0.121)
Inflation X prodgap t-1
0.39*** 0.05 0.23 -0.07
(0.135) (0.132) (0.142) (0.207)
N 2,492 2,492 2,492 2,492 2,492 2,492 2,492
R2 0.087 0.100 0.100 0.091 0.110 0.083
AR(1)
0.000
AR(2)
0.137
Sargan
0.000
Number of instruments
335
Number of regions 265 265 265 265 265 265
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
52
Table 8: Robustness checks for the macroeconomic variables
Dependent variable:
regional productivity
growth
Removing
frontier regions Removing top 3 regions
Alternative frontier region -
removing top 3
(1) (2) (3) (4) (5) (6)
FE
System
GMM FE
System
GMM FE
System
GMM
Frontiergrowth 0.06*** 0.07*** 0.06*** 0.07*** 0.28*** 0.27***
(0.019) (0.021) (0.020) (0.022) (0.038) (0.043)
Productivity gap t-1 -1.81 5.71 -1.09 5.68 -14.52*** -2.83
(2.114) (3.896) (2.655) (4.560) (3.120) (6.392)
Trade t-1 -0.24*** -0.45*** -0.26*** -0.52*** -0.07 -0.37***
(0.049) (0.086) (0.067) (0.099) (0.064) (0.102)
Trade X prodgap t-1 -0.17*** -0.52*** -0.20** -0.63*** -0.11 -0.87***
(0.055) (0.142) (0.085) (0.135) (0.104) (0.191)
Debt t-1 0.04** 0.05*** 0.05*** 0.07*** 0.05*** 0.06***
(0.015) (0.019) (0.017) (0.022) (0.016) (0.021)
Debt X prodgap t-1 -0.01 -0.00 -0.00 0.02 -0.00 0.01
(0.008) (0.014) (0.009) (0.014) (0.011) (0.019)
Inflation t-1 -0.10 -0.04 -0.11 0.08 -0.07 0.09
(0.092) (0.157) (0.125) (0.186) (0.097) (0.151)
Inflation X prodgap t-1 -0.05 -0.10 -0.07 0.07 0.01 0.22
(0.146) (0.265) (0.180) (0.288) (0.203) (0.350)
N 2,281 2,281 1,856 1,856 1,856 1,856
R2 0.125
0.129
0.206
AR(1)
0.000
0.000
0.000
AR(2)
0.181
0.169
0.259
Sargan
0.000
0.000
0.000
Number of instruments
319
286
286
Number of regions 249 249 216 216 216 216
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
53
Table 9: Further robustness checks for the macroeconomic variables
Dependent variable: regional
productivity growth Without Spain Without Poland Without Ireland
(1) (2) (3) (4) (5) (6)
FE
System
GMM FE
System
GMM FE
System
GMM
Frontiergrowth 0.05** 0.05** 0.05** 0.06*** 0.05*** 0.06***
(0.020) (0.021) (0.019) (0.020) (0.019) (0.020)
Productivity gap t-1 -2.17 3.33 -2.33 2.07 -2.13 3.08
(2.035) (3.614) (1.948) (3.455) (1.986) (3.637)
Trade t-1 -0.17*** -0.29*** -0.14*** -0.26*** -0.20*** -0.33***
(0.045) (0.073) (0.046) (0.076) (0.045) (0.075)
Trade X prodgap t-1 -0.12** -0.38*** -0.12** -0.37*** -0.13*** -0.40***
(0.049) (0.109) (0.047) (0.113) (0.050) (0.113)
Debt t-1 0.02 0.02 0.02* 0.03* 0.02* 0.03*
(0.014) (0.019) (0.012) (0.016) (0.013) (0.016)
Debt X prodgap t-1 -0.02** -0.02 -0.01* -0.01 -0.01* -0.01
(0.008) (0.013) (0.008) (0.012) (0.008) (0.013)
Inflation t-1 -0.01 0.02 -0.03 0.11 -0.05 -0.01
(0.079) (0.121) (0.080) (0.128) (0.079) (0.124)
Inflation X prodgap t-1 0.08 -0.04 0.14 0.18 0.03 -0.08
(0.132) (0.206) (0.132) (0.212) (0.132) (0.211)
N 2,309 2,309 2,355 2,355 2,474 2,474
R2 0.095
0.107
0.112
AR(1)
0.000
0.000
0.000
AR(2)
0.245
0.135
0.131
Sargan
0.000
0.000
0.000
Number of instruments
316
319
333
Number of regions 246 246 249 249 263 263
Dependent variable is regional productivity growth at the TL2 level. Robust standard errors in parentheses. Standard errors are clustered by region. ***, **, * indicate significant at the 1%, 5% and 10% level respectively.
54
Appendix
Fixed effects estimates of marginal effects of the macroeconomic variables:
Figure A1: Marginal effects of macroeconomic variables, OLS.
55
Figure A2: Excluding frontier regions, OLS.
Figure A3: Excluding top three regions, OLS.
56
Figure A4: Alternative frontier region, OLS.
Figure A5: Excluding Spanish regions, OLS.
57
Figure A6: Excluding Polish regions, OLS.
Figure A7: Excluding Irish regions, OLS.